You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+55-4Lines changed: 55 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -7,19 +7,70 @@ This repository uses both Pytorch and Tensorflow Serving (TFS).
7
7
8
8
# Installation
9
9
10
-
Both Google Colab and singularity environments are supported. This environment is used because it is a convenient method to have both pytorch and tensorflow present.
10
+
## Runtime Environments
11
11
12
-
## Singularity Containers
12
+
This repository supports both Docker and Singularity environments.
13
13
14
-
See the [container definition file](vm/deployment-runtime-RHEL9.def) in the vm folder. This container is based off a google colab public docker.
14
+
The dockerfile is provided at the root of the repository ([Dockerfile](Dockerfile)), and the singularity
15
+
definition file is in the `vm` folder ([singularity.def](vm/singularity.def)).
16
+
17
+
To learn more about how we support this, please read [vm/README.md](vm/README.md).
18
+
19
+
## Development
20
+
This repository uses [uv](https://uv.run/) to manage multiple python environments.
21
+
To install uv, see the [uv installation instructions](https://uv.run/docs/installation).
22
+
23
+
To create the development environment, run:
24
+
```
25
+
uv sync --extra cpu
26
+
```
27
+
28
+
If you happen to have access to a GPU, you can create a GPU-enabled environment with:
29
+
```
30
+
uv sync --extra gpu
31
+
```
15
32
16
33
# Available Models
17
34
18
35
See [model docs](docs/models.md) for information about available models.
19
36
37
+
## Model Directory Configuration
38
+
39
+
The model directory can be configured at runtime using environment variables or nextflow parameters:
40
+
41
+
### Environment Variable
42
+
Set the `MOUSE_TRACKING_MODEL_DIRECTORY` environment variable:
Use the `--model_dir` parameter when running nextflow:
49
+
```bash
50
+
nextflow run main.nf --model_dir /path/to/your/models --input_batch video_batch.txt --workflow single-mouse
51
+
```
52
+
53
+
### Default Location
54
+
By default, models are expected at `/kumar_lab_models/models/`. The directory structure should follow:
55
+
```
56
+
models/
57
+
├── pytorch-models/
58
+
│ ├── single-mouse-pose/
59
+
│ ├── multi-mouse-pose/
60
+
│ └── fecal-boli/
61
+
└── tfs-models/
62
+
├── single-mouse-segmentation/
63
+
├── multi-mouse-segmentation/
64
+
├── multi-mouse-identity/
65
+
├── static-object-arena/
66
+
├── static-object-food/
67
+
└── static-object-lixit/
68
+
```
69
+
20
70
# Running a pipeline
21
71
22
-
Pipelines are run using nextflow. For a list of all available parameters, see [nextflow parameters](nextflow.config). Not all parameters will affect all pipeline workflows.
72
+
Pipelines are run using nextflow. For a list of all available parameters, see
73
+
[nextflow parameters](nextflow.config). Not all parameters will affect all pipeline workflows.
23
74
24
75
You will need a batch file that lists the input files to process.
0 commit comments